SSD vs. HDD: Show Me the Numbers

I’ve done a lot of talking about how SSDs1Solid State Drives are faster than HDDs,2Hard Disk Drives but I think it’s time for some cold, hard benchmarks.  Cue the fight announcer voice!

IIIIIIIIIN THE RIGHT CORNER, we have the latest of from a long lineage of magnetic storage, the protector of Joe’s Documents and Games, and the reigning default of computer storage for nearly 3 decades, give a hand to the HDD!

AAAAAND IN THE LEFT CORNER, we have the newcomer, finally making the jump from the world of high-bandwidth servers down to us consumers, and champion of Joe’s OSes and Programs, please welcome the SSD!

Contestants Ready?  Fight!

SSD on the left, HDD on the right.

That … wasn’t even close.  The SSD squarely beats the HDD in every speed measurement.  Let’s break down what each of these numbers mean.

Sequential Read

The first row is sequential read/write speeds, that is, data stored all in a row.  As you can see, the SSD is ~2.3x faster than the hard drive.  Remember that this kind of task is the HDD’s best case scenario: it can read data without moving the mechanical head too much.  But it still has to wait for the data to come around on the spinning disk, while the SSD is ready to read the data it wants whenever it wants.

Random Reads

The next 3 rows are all about random read/write speeds, that is, data stored all over the place.  If you want to read the data, you have to find each part first.  Since the SSD doesn’t have to go searching for the data, unsurprisingly, it reads/writes nearly as well as it does sequentially.  This is where hard drives really lose; all the movement adds more and more time to the action, and just absolutely chokes on the task.

Now you may be wondering, why are there 3 different measurements of random reads?  Honestly?  I really don’t know haha.  From what I can gather, it has to do with the number of reads that the computer lines up at a time.  For example, the Q32 test basically lines up 32 different requests for data at the same time whereas Q1 is just one request.  What does that actually test?  Yeah, I don’t know, and I haven’t found anything about why I should care, so meh.

Speed discrepancy?

Now, if you’re like me, you’re looking at the SSD and HDD sequential read/write speeds, and are amazed that the HDD was kind of close.  I mean, yeah, the SSD is still more than twice as fast than the HDD, but that still seems kinda close?  What gives?  Well, it has to do with the communication interface I’m using.  I have both drives connected to my computer’s SATA (Serial ATA3yeah, I’m not kidding, look it up.  It’s the updated version of PATA, ie Parallel ATA, ie the thick ribbon cables from the 90s) ports.  But my desktop computer is from 2010-ish.  Which means that my ports are SATA rev. II, which is limited to 300 MB/s.  And you can see that the benchmarks of 267/261 MB/s are pretty close to that theoretical maximum.4The difference is likely due to 1) overhead and 2) it’s my main boot drive, so the OS was pinging it during the whole test  So my SSD is probably capable of faster transfer speeds, but is limited by my other hardware.

This brings up another point.  All computers today will have at minimum the SATA rev. III standard, which is twice as fast as my computer’s.  And if that wasn’t enough, manufactures have also started building drive communication standards based on the PCIe peripheral standards that gives a direct line from the drive to the CPU with minimal overhead.  These drives have truly mind-boggling performance, but at a pretty mind-boggling price.

If you didn’t catch much of that, don’t worry.  Suffice it to say the difference between my HDD and SSD feels like night and day, and newer SSDs and computers will be even better.  But if you want more information, Linus Tech Tips has a pretty complete video that goes through a bunch of the physical and protocol standards that show up in a lot of laptops today.

How representative are those numbers really?

I mean, come on, the SSD is measuring more than 256x faster than the HDD in the random read test.  That’s way too good to be true!

Well, remember to take the whole picture into consideration.  Remember back in my HDD post, I mentioned that Windows Vista and beyond automatically defragmented your hard drive?  Because of that optimization, the HDD realistically doesn’t have to jump back and forth as frequently as the random test benchmarked here.  But, defragmenting isn’t perfect, and the head and the platters definitely still have to move, causing latency.  Realistic, day-to-day usage will definitely land somewhere between the two categories, probably closer to the sequential read numbers.

In my own (non-empirical) experience, my Notes Laptop took 5-7 minutes to load MATLAB5a giant data analysis program I use on a daily basis while it still had a HDD, but now only takes about 30s – 1min.  Still not particularly quick, thanks to a hopelessly outdated CPU, but leaps and bounds better than it was.  It was impractical to use MATLAB on that machine before, but now I can realistically use it to generate some figures during a meeting if I need to.

So, no, your computer is not going to be 256x faster after you upgrade to a SSD.  But I’d say from experience 3-5x faster is about right.

tl;dr

Finally, some real numbers to back up the handwavey nonsense from before.

My desktop’s SSD is ~2x faster in sequential reads and ~250x faster in random reads than my desktop’s HDD.  That translates to about 3-5x faster in real-world usage.

Due to the age-related technology limitations of my 8-year-old desktop, a new laptop’s speeds should be even faster.

I really should write a post about the computers I have, since I keep referencing them.

Leave a Reply

Your email address will not be published. Required fields are marked *